Estimation of Two Projection Indices by Wavelet Kernel Functions
Dang Huai-yi · 2007
The key procedure of exploratory projection pursuit is the optimization of a criterion function called the projection pursuit index. In this paper,the wavelet kernel estimations for two kinds of projection indices,the Cook index and PPDA index,are proposed. We establish the properties of wavelet kernel-based projection indexes,including the asymptotic unbiased and the convergence in mean squared sense. Some results of projection indexes with wavelet kernel-based estimations are compared with that of the Gauss kernel estimation,which indicate the e?ciency of the wavelet kernel-based estimations.